An obstacle in realizing next-generation microelectronic and optoelectronic devices and optimal integration of these devices is found in lattice mismatches between different crystals of group III-V semiconductor materials. Generally, the lattice mismatch between a substrate and an epitaxial over-layer induces strains within the over-layer. This may lead to strain relaxation which can result in formation of material defects such as dislocations within the crystalline structure of the over-layer. FIG. 1 illustrates a mismatched over-layer 1 epitaxially grown over a substrate 2, the boundary between the over-layer 1 and the substrate 2 being indicated with reference numeral 4. As shown in FIG. 1, the lattice constant associated with the over-layer 1 is different from the lattice constant associated with the substrate 2, hence the term “mismatched over-layer”. Strain relaxation due to lattice mismatch is accommodated by the formation of mismatch dislocations 3 within the crystal. Defects within a crystal generally degrade the performance of devices made from the crystal, because such defects can scatter movement of carriers (electrons and holes) and can act as carrier traps and/or recombination centers. It is thus useful to provide means for growing a crystal over-layer which has different lattice constant from the substrate on which the over-layer is grown, in such a fashion that strain relaxation does not occur and mismatch dislocations do not form. FIG. 2 is an example of this, in which the structure of over-layer 1 is preserved and no mismatch dislocations are formed.
In the prior art, two main approaches are used to address the lattice mismatch problem and the strain relaxation it causes:                1) In a first approach, defects are confined in thick relaxed buffers so that the top active layer of a device can be of a different lattice constant from that of the substrate and is as defect free as possible.        2) In a second approach, thin compliant solid layers are bonded to foreign substrates and re-growth is performed.        
However, these approaches still present performance degradation problems. A buffer layer of defects degrades the quality of the active layer on top of the buffer layer used for a device. In addition, thick buffer layers are not very suitable for device fabrication because high mesa or deep isolation implants are then necessary for device isolation, and can result in high leakage currents and low wafer yields. Further, procedures for implementing the second approach are rather complicated due to problems associated with wafer bonding, fabrication of thin layers (tens of Å in thickness) and re-growth on surfaces contaminated in the wafer-bonding and fabrication processes.
Hence, there is a need for a method of growing a crystal over a substrate such that mismatch dislocations are prevented from appearing within the crystal, even though the crystal and the substrate have different lattice constants.